In a hurry?
The quickest tests are behavioural, not visual. Ask for a short video call with a specific gesture, and check whether their story stays consistent. A profile that dodges both is fake. If money comes up, it is a scam, full stop. Report it: Report a scam
Start with the honest number
Industry analysis of over eight million profiles puts fake accounts at roughly one in ten of all new dating profiles created. Our own network runs at TODO: % of new registrations removed as fake, trailing 12 months, from TODO: profiles reviewed, trailing 12 months profiles reviewed across more than 1,000 communities, with TODO: % blocked at signup stopped before a genuine member ever saw them. We publish these numbers because most of the industry will not, and because the fakes that reach you are the ones that got past everyone's filters, including ours. This guide is about catching those.
The 2026 problem: your eyes are no longer enough
For years, the advice was to squint at the photos. Mangled teeth, warped ears, smudged backgrounds: the fingerprints of early AI images. That advice is now out of date, and repeating it gets people hurt. Research published this year found that people who rate themselves good at spotting AI faces performed barely better than a coin flip against current generators. The obvious flaws are gone.
Two useful weaknesses remain. First, current AI portraits trend towards an eerie averageness: perfectly symmetrical, studio-lit, frictionless faces, every photo a polished solo shot. Real profiles are messier, with bad angles, group shots and photos where the subject is not the point. A profile that is all portrait and no life is a flag. Second, AI still struggles to generate the same fictional person consistently across many photos and situations, and it cannot produce a photo on demand. That second weakness is the one you can use.
Photos: stolen or generated, and what each means
Fake profiles use one of two photo strategies, and they fail differently.
Stolen photos, the classic catfish, appear elsewhere online, which is why the reverse image search still earns its place. Our photo checker matches images against pictures our moderation team has already removed, and the same stolen portrait routinely resurfaces across profile after profile; taking down one fake rarely means taking down the only copy. One search ends the doubt. Check a photo
AI-generated photos return no matches anywhere, and scammers adopted them precisely because they beat reverse search. TODO: % of removed fake profiles using AI-generated imagery, and the 12-month trend of the fake profiles we remove now use generated imagery. So learn to read the no-match result correctly: on a brand-new profile with studio-perfect solo photos, finding nothing is not reassurance. It is a yellow flag of its own.
The tests that still work
The specific-gesture test. Ask for a quick video call, or a selfie doing something you choose: three fingers up, touching their left ear, holding today's newspaper if you like tradition. Real people find this mildly odd and do it anyway. Generators and deepfakes cannot reliably match an arbitrary request in real time, and scammers know it, so what you get back is an excuse. One excuse is life. A pattern of excuses is your answer.
The consistency test. Scammers run many conversations at once, usually from scripts. Revisit something they told you last week, slightly wrong, and see what happens. Real people correct you. Script-runners agree with your wrong version, because they never knew the right one.
The clock test. Instant, polished replies at every hour, endless availability, and empathy that never misfires read as attentiveness and are actually automation. Real people sleep, miss messages, and occasionally respond to the wrong part of what you said.
The footprint test. Most real people leave some trace online that matches their story. No trace proves nothing on its own, some people are just private, but no trace plus perfect photos plus a rushed connection is a pattern, and patterns are how our own systems catch them: TODO: % of fake profiles displaying 3+ tells from the standard cluster.
Catfish or scammer? The difference is the ask
Not every fake profile wants your money. A catfish in the original sense is usually a real, lonely person hiding behind someone else's face, and the harm is emotional. A scam profile is a business, and it always, eventually, arrives at money, an "investment", or a request that can be monetised.
You do not need to diagnose which you are talking to. The response is the same, because the test is the same. If someone you have never met in person asks you for money, it is a scam. Not sometimes. Not "but this feels different". Every time. And if there is no money ask but they will not verify with a live call after weeks of intensity, you are in a relationship with a mask, whatever is behind it.
What to do
Do not announce your suspicions; you will only sharpen the next attempt. Stop investing, run the photo check, apply the gesture test, and report the profile whether or not you are certain. On the communities our team moderates, member reports are a detection signal in their own right: TODO: % of fake-profile removals initiated or corroborated by member reports. You reporting one profile takes down the copies of it you never saw.
If money has already moved, bank first, report second, in every country: Get help in your country